Author

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 315. (Read 3426976 times)

hero member
Activity: 840
Merit: 1000
please nobody release a GPU miner for cryptonite. Leave us alone with a coin that is profitable on CPU. I really mean this. Let us utilize every processor in a PC rather than choosing which coin to mine on GPU while CPU sits idle.

There's already an AMD/ati miner.

 That is sad. but I am still making some money on cryptonite.
newbie
Activity: 59
Merit: 0
please nobody release a GPU miner for cryptonite. Leave us alone with a coin that is profitable on CPU. I really mean this. Let us utilize every processor in a PC rather than choosing which coin to mine on GPU while CPU sits idle.

There's already an AMD/ati miner.
hero member
Activity: 840
Merit: 1000
please nobody release a GPU miner for cryptonite. Leave us alone with a coin that is profitable on CPU. I really mean this. Let us utilize every processor in a PC rather than choosing which coin to mine on GPU while CPU sits idle.
hero member
Activity: 840
Merit: 1000
@QuadroQ

Also with blake-256 My guess is that cudaminer uses too much pcie bandwidth because I get 290mh/s with 1 card it becomes 270-280 with 2 cards and with 3 cards hashrate for only 2 cards r shown in consoleand 3rd card's hashrate is shown only when new block is found(40-60 mh/s) or a share is accepted (total hashrate around 820-850 Mh/s). So if you r using more than 1 card it will not scale smoothly.
hero member
Activity: 644
Merit: 500
And a little update.

The good: Managed to squeeze out a decent ~40% buff to hashrate, now pulling 330 W instead of 270 W. Less frequently accessed stuff in global memory and vice versa, no magic there.

Code:
[2014-06-24 07:31:42] accepted: 7/7 (100.00%), 1323.63 H/s (yay!!!)
[2014-06-24 07:31:42] GPU #5: GeForce GTX 750 Ti, 226.77 H/s
[2014-06-24 07:31:42] GPU #3: GeForce GTX 750 Ti, 226.87 H/s
[2014-06-24 07:31:43] GPU #4: GeForce GTX 750 Ti, 226.51 H/s
[2014-06-24 07:31:43] GPU #1: GeForce GTX 750 Ti, 229.07 H/s
[2014-06-24 07:31:44] GPU #0: GeForce GTX 750 Ti, 228.27 H/s
[2014-06-24 07:31:44] GPU #2: GeForce GTX 750 Ti, 227.61 H/s

The bad: The kernel still runs stupid slow and eats up Windows desktops for breakfast, lunch and dinner. And snacks. Lots of snacks, practically nonstop. I did find a launch config with nice balance between hashrate and kernel run time, keeping it below the 2 second TDR timeout and hitting about 200 H/s but it still takes way too long and still pretty much destroys desktop interactivity. Basically run it on another card in your system and leave the one with your display alone or walk away from the computer after starting the miner.


Nice work!
I'm already running it from my iGPU ^^" I don't really care, I can always switch back when gaming Wink Can't wait though Smiley
full member
Activity: 137
Merit: 100
And a little update.

The good: Managed to squeeze out a decent ~40% buff to hashrate, now pulling 330 W instead of 270 W. Less frequently accessed stuff in global memory and vice versa, no magic there.

Code:
[2014-06-24 07:31:42] accepted: 7/7 (100.00%), 1323.63 H/s (yay!!!)
[2014-06-24 07:31:42] GPU #5: GeForce GTX 750 Ti, 226.77 H/s
[2014-06-24 07:31:42] GPU #3: GeForce GTX 750 Ti, 226.87 H/s
[2014-06-24 07:31:43] GPU #4: GeForce GTX 750 Ti, 226.51 H/s
[2014-06-24 07:31:43] GPU #1: GeForce GTX 750 Ti, 229.07 H/s
[2014-06-24 07:31:44] GPU #0: GeForce GTX 750 Ti, 228.27 H/s
[2014-06-24 07:31:44] GPU #2: GeForce GTX 750 Ti, 227.61 H/s

The bad: The kernel still runs stupid slow and eats up Windows desktops for breakfast, lunch and dinner. And snacks. Lots of snacks, practically nonstop. I did find a launch config with nice balance between hashrate and kernel run time, keeping it below the 2 second TDR timeout and hitting about 200 H/s but it still takes way too long and still pretty much destroys desktop interactivity. Basically run it on another card in your system and leave the one with your display alone or walk away from the computer after starting the miner.
sr. member
Activity: 480
Merit: 250
Forgive me if this has already been brought up (I can't read through all 800 pages), but ccminer seems to be missing two very popular algo's that would pretty much make it complete: CryptoNight (Monero), and Blake-256 (many coins and growing). I know that CudaMiner has a Blake-256 implementation, but it's performance is so abysmal that it's basically useless (which is surprising considering the simplicity of the hashing algorithm). I've been very happy with the X11/X13 performance, and would love to see equally good implementations of these two profitable algo's. Smiley

I'm assuming that the newest version of CUDAminer has the best version Cbuchner made of Blake-256.
As for CryptoNight, someone posted screenshots of a version that (s)he made that runs it, but that version is not available to the public.

Well if anyone has any experience using CudaMiner with Blake-256, perhaps you could explain what you're doing to get it to work, and what kind of hashrate you're getting. I was getting terrible hashrates and the pool wouldn't accept any of my shares at all when I tried it.

So CryptoNight is on it's way, but not publicly available yet?

I have used cudaminer to mine blake-256 coins. You are right performance is abysmal compared to anything else. Firstly here is the string I am using for mining.

cudaminer.exe -i 0 --algo=blake -o stratum+tcp://eu3.blakecoin.com:3334 -O user:pass -L 512 -l T2304x24

I havent tweaked with all the parameters and autotune but this gives me around 290 mh/s per 750ti.

Christian said that this is not a optimized version for blake-256 so it runs but abysmally. He is going to add blake-256 to ccminer later (I hope you remember this and going to release it soon Christian Sad )

Also you cant get any shares because all pools are set at fixed diff of 32 which is very high for hashrate belo 1.5 gh/s. So you have to wait for long times between shares but in the long run it evens out. I get some 125-140 shares in 18-20 hours.

I hope this helps. If you are able to get any better hashrate then please share with me and others. You can see configs and submit them at cudmining.cc

I appreciate your reply, and your confirmation that Blake-256 performance is very poor in it's current implementation. I figured it must not be optimized since the algo is relatively simple (especially compared to X11/X13). I believe there's a chance that nvidia could steal the Blake-256 crown away from AMD, with a good implementation, so I second your reminder to Christian!

As I said originally, if ccminer get's a solid implementation of the CryptoNight and Blake-256 algorithms, then other than some performance tweaking, it'll cover all the bases (currently).

In my dreams I'd love to see a miner that combines the sgminer/cgminer/cudaminer/ccminer/cpuminer into one application with a unified interface (similar to sgminer, but better). Maybe one day.
hero member
Activity: 840
Merit: 1000
Forgive me if this has already been brought up (I can't read through all 800 pages), but ccminer seems to be missing two very popular algo's that would pretty much make it complete: CryptoNight (Monero), and Blake-256 (many coins and growing). I know that CudaMiner has a Blake-256 implementation, but it's performance is so abysmal that it's basically useless (which is surprising considering the simplicity of the hashing algorithm). I've been very happy with the X11/X13 performance, and would love to see equally good implementations of these two profitable algo's. Smiley

I'm assuming that the newest version of CUDAminer has the best version Cbuchner made of Blake-256.
As for CryptoNight, someone posted screenshots of a version that (s)he made that runs it, but that version is not available to the public.

Well if anyone has any experience using CudaMiner with Blake-256, perhaps you could explain what you're doing to get it to work, and what kind of hashrate you're getting. I was getting terrible hashrates and the pool wouldn't accept any of my shares at all when I tried it.

So CryptoNight is on it's way, but not publicly available yet?

I have used cudaminer to mine blake-256 coins. You are right performance is abysmal compared to anything else. Firstly here is the string I am using for mining.

cudaminer.exe -i 0 --algo=blake -o stratum+tcp://eu3.blakecoin.com:3334 -O user:pass -L 512 -l T2304x24

I havent tweaked with all the parameters and autotune but this gives me around 290 mh/s per 750ti.

Christian said that this is not a optimized version for blake-256 so it runs but abysmally. He is going to add blake-256 to ccminer later (I hope you remember this and going to release it soon Christian Sad )

Also you cant get any shares because all pools are set at fixed diff of 32 which is very high for hashrate belo 1.5 gh/s. So you have to wait for long times between shares but in the long run it evens out. I get some 125-140 shares in 18-20 hours.

I hope this helps. If you are able to get any better hashrate then please share with me and others. You can see configs and submit them at cudmining.cc
full member
Activity: 137
Merit: 100
Excellent news, Sir! We managed to be faster. But I don't think I should publicly state by how much...

Let me know if you want the vacant position as a co-author on ccMiner Wink

Did you have to merge in some of the changes from https://github.com/LucasJones/cpuminer-multi to get the protocol support
for cryptonight (e.g. JSON RPC v2)?

Christian

C&C miner faster than the trained-monkey-mashing-on-keyboard version? I'm shocked  Tongue

And you're dead on about JSON-RPC 2 and where the code for it came from.
sr. member
Activity: 602
Merit: 250
I have 16 750 Ti's at my disposal to help beta test and get this pool off the ground (and growing). Let me know if I can be of assistance.
full member
Activity: 238
Merit: 100
Medichain: The Medical Big-Data Platform
I guess I am not, I don't even understand about what you are talking about  Grin
(speaking of phone, for a few days, I had to mine through my mobile phone (used as a wifi hotspot...) and actually it worked quite well...

I did this as well when we had an internet outtage a few weeks back.

Turned on the hotspot and it was mining away.
sr. member
Activity: 364
Merit: 250
Thankyou for the detailed answered, so in essence it's just like how profit-switching multipools work, probably closer to the model that TMB uses with the proxy stratum for different algos. They also proxy to a bigger pool if they don't have enough hash to generate blocks in timely manner, or used to anyway.

TMB???

Who are they?  Not ringing a bell.
TradeMyBit
full member
Activity: 168
Merit: 100
Thankyou for the detailed answered, so in essence it's just like how profit-switching multipools work, probably closer to the model that TMB uses with the proxy stratum for different algos. They also proxy to a bigger pool if they don't have enough hash to generate blocks in timely manner, or used to anyway.

TMB???

Who are they?  Not ringing a bell.
full member
Activity: 168
Merit: 100
Gotten some questions via PM so thought I'd head them off in the thread if possible.
1) Can you start testing it now?
A) Soon, if you are interested in being involved in alpha/beta testing before it goes public send me a PM with your rig details.  I'll start to open this up in a few days to additional miners.

2) Can you get coin X instead of BTC (ie TAC, or XMR)?
A) Let's call it a phase 3 effort.  It's on the todo list but probably won't be available until a month from now.  However ideally if for example you wanted to get paid in TAC and we mined 500 TAC while you earned the equiv of 50 TAC then sure we would just pay you the 50 TAC while taking the remaining 450 to market.  This benefits the whole group as there will be less fees overall. You would have saved on fees to convert to btc and of course any fees for you to take your btc and buy the TAC.  Substitute any coin here for TAC.  This option will only be available if it's a coin we mined that day.  But this could become part of the calculator used to determine what to mine as it would reduce fees.

3) Will it support AMD cards?
Yes as already stated it will support basically AMD, nVidia or CPUs.  If you look at the "windows form" in the screen shot I provided earlier you can set the working directory and program to use for each algo.  Not such a big deal for nVidia hardware as you will either use ccminer or Cudaminer but for AMD or CPU mining you could have a different directory/program setup for each algo.

4) How do we know this isn't a scam and you will keep all the funds?
A)  Sad, but valid question these days.  Unless you know me, a only a few do, you don't. Besides the fact I use my real name in the forums (fist initial last name) and that I sign most messages with my first name you pretty much know my full name. Smiley My best advise is sit back and wait for others to try first, or only try it with a GPU or two so you aren't out much of anything if I'm scamming you.  Sadly I'm pretty smart and wouldn't go to such troubles to scam you. If I wanted to scam you I could do it much easier then this. Smiley

5) Is this going to be open source or closed source?
Really hard to answer.  The backend is all custom programming for my environment.  Not really hard for a .NET programmer used to working with IIS, SQL server and C#.  Not LAMP for sure!  The backend I'd consider proprietary and "closed source" at this time.  The front end is really simple and could be made open source so people know what software is running on their computer.  This could also help with a Linux version.

6) Will there be any forced "donations" involved?
A) At present NO.
However, I have sent PM to Chris about this specifically in regards to BBR or CryptoNight.  I'm totally open to this.  If for example the best we can do is 0.01 btc per day for a rig of 3- 750ti (just made up number) but Chris has a "killer" algo that could do 0.003 on the same rig then I think either a fixed amount of "donation" or "time slice" could be given to him.  This would benefit all parties as everyone makes more $ or btc.

For now, NO, but I'm open to it if both Chris and I could develop a method that works for both of us.

7) What kind of fee will I charge for this service?
A) At present 0%.  During Alpha, Beta cycle there will be no fee.  It's possible you could lose funds using my service during the first week or so as it will be using hash to generate stats used to figure out the best coins to mine (during alpha/beta).  However, I guarantee, I'll make it up to you. Smiley

This is also why I don't want to roll this out prematurely before I get a few people who can help me test it.  IE part of your hash through me and part the same way you have been mining.
But if you lose any funds it won't be much.  You would also have to track everything pretty tightly yourself already to do better.

But to answer your question. I don't know.  It could be 1 or 2% or 15 to 30 minutes a day of your mining time.  I will need to evaluate what the costs are vs the income to determine this properly.  Depend on total hash rate of course. I'll be upfront about it and of course only if this is really more profitable then what you could do on your own.

True costs are hard to determine at present.  The pure hardware and monthly expense will be easy to calculate but the "human factor" is not known yet. As I've previously mentioned I'd probably want to to get a few others (trusted souls) involved in the project so there is 24 hours of "eyes" on the backend as well as 24 hours of trading "expertise" on hand to maximize our exchange trades.

So this is a question I can't really answer at present.  More expertise on hand 24 hours should get better results but it's hypothetical at this point, so I don't know.

I think that covers the questions I've gotten thud far.

Carlo
sr. member
Activity: 364
Merit: 250
legendary
Activity: 1400
Merit: 1050
I don't think the Christian version of the Cryptonite miner has this problem. Maybe we could try that one out  Cool Kiss

It's too fast, it burnt out one of my ASUS ROG MARS cards. Wink


Now im unsure if this is true or not haha

On a side note, if anyone here is a phone geek and has an invite sitting around for the oneplus one phone please pm me
I guess I am not, I don't even understand about what you are talking about  Grin
(speaking of phone, for a few days, I had to mine through my mobile phone (used as a wifi hotspot...) and actually it worked quite well...
full member
Activity: 238
Merit: 100
Medichain: The Medical Big-Data Platform
What world are we living in when mining bitcoin shows up as the most profitable on CoinWarz?

WTH?!

full member
Activity: 168
Merit: 100
So CryptoNight is on it's way, but not publicly available yet?

tsiv is having a working prototype of Cryptonite it appears. But it seems it is not yet quite ready for prime time, as it will nearly freeze the machine while mining and cause windows driver timeouts unless you apply registry hacks. For a release this will have to be a bit more polished.



I don't think the Christian version of the Cryptonite miner has this problem. Maybe we could try that one out  Cool Kiss

Yea but I heard that Christian's miner only does 30H/s compared to tviv's which does 160H/s.

Christian can proved me wrong with a non-photoshopped screen shot.   Shocked
full member
Activity: 168
Merit: 100
At present all payments would be made to a Bitcoin address.
How come?

Is it because users will be limited only to pools of your choice when determining the most profitable coin?
Short question, really, really long answer! Smiley

Yes and No.  Let me explain.

I would control the pools used.  Pools would be setup as needed based on profitability of different coins.  In a nutshell there will be pools setup for the following algorithms:
CryptoNight, X11, JHA, NIST5, JHA. X13, Groestl, etc  Each of these would be assigned a port number.

As an Extreme example lets just say there are 25 people using the service with 2 rigs each for a total of 50 rigs and lets say 200 GPUs.  Now lets also for the sake of argument assume it's a strange day and profit is changing a lot between different coins (don't care if same algo or not).  

We could mine some x11 Millionaire, some x13 Marucoin, some JHA jackpot and some nist5 TAC.  Now if you only have a rig or two you aren't going to be making a ton to begin with and I'm sure you don't want to have to 4 wallets and do trades for 4 different coins in a timely manner to capitalize on the trends..  So my thought is that I'd use/leverage the full power of the 50 rigs to maximize revenue and trade the 4 coins at the appropriate time for the "group".  This could mean cashing in right away or holding off a bit because the coins already mined are rising in value at an exchange.

Basically convert everything to BTC.  Then using the accounting of shares of each coin mined/traded I could calculate your fair share and send you one btc payment once to a few times a day depending on your volume. (super high level explanation).

Where things will get interesting is that I will be able to take into consideration the time for a block to mature to get it to market to sell. If the coin is going up in price it's worth the gamble, and if the coin is profitable but going down in price or doesn't have the needed volume or block confirmation time is long we can pass on it. The backend can take into consideration trading volume and how much of a "buy wall" is present. Things of this nature. Due to programming already in place the backend will learn what the fudge factor is for each coin and will continually look at it and refine the process.

Also I'll integrate with services like NiceHash (and similar) to be able to rent our MH on demand if it's more then we can presently mine for. We always want to earn the most we can per MH of hash so...

To start for the most part it will just use the "pure" calculators I've built to figure out what is the most profitable coin.  But already built in (without the data) is a "fudge" factor for each coin, algo or pool used.  So as an example lets say a particular coin by pure math should produce 100 coins a day.  We all know it will never be 100 coins long term (luck aside) due to things like orphon blocks, stale shares, network latency, DDos attackes, etc.  So for each coin/algo/provider the system will quickly get smart and learn that coin A only does 95% of calculated amount.  Coin B does 90%, coin C 85%, coin D does 96%. As time goes on it will learn to use the "fudge factor" in determining the "real" most profitable coin.

So in a nutshell I will be able to use the massive MH rate we can all produce as a group to analyze and produce code that gets us the most $ for the hash we have combined.

Now where we will differ from tradition pools is that I will not run a typical pool per say. In other words it's not going to be like "solo" mining with the 50 rigs or 200 GPUs.  I'll be doing a "proxy stratum" for each algo.  So if you are mining X11 you'll point to port 6001 (example). This could in turn use a pool or two running the coin we want to mine.  The advantage of this is that if our "calc" determine we should switch off to JPC (jackpot) then all the shares we have already submitted for X11 will still be good and payable soon. Since we are mining at already profitable pools the income stream should be quick and blocks should not take long. Since we are using one or more other pools for that X11 coin and since all their hash rate didn't go "poof" when we switched they will continue to mine and will solve the block.  We will get paid for our shares.  In contrast to many "multi-pools" they switch coins and you "loose" your shares in that coin until they switch back to it if ever!!!

To begin we'll just appear as a normal user to these pools with massive volume per our account. As we grow I'll try and negotiate with different providers for better rates on our behalf.

Now one of the cool things is that due to the "buy wall" we could have some people mining X11 and others mining JPC while yet others are getting some NiceHash X11.  This will all be taken care of by the backend to "normalize" the income.  So it won't matter to you if you are mining a coin directly on X11 or via a "rental/hash" company.  You will/should make the most available.

Getting back to your original question now.  With all the stuff that can/will be happening on the backend it will be easiest to "normalize" everything based on shares to a common demonstrator which in this case is BTC.  Hence your "login username" will just be a btc address.

Make sense?

Carlo

PS some of the backend is just "brain power" or gut instinct on when to sell/trade, hold or sell for most $.  It's foolish to think it can completely be automated. People have tried that for years and years on the stock market.  Because of this and since I can't do this all myself I do plan on "recruiting" a few trusted soles into the "mix" to help run things.

legendary
Activity: 1400
Merit: 1000
I don't think the Christian version of the Cryptonite miner has this problem. Maybe we could try that one out  Cool Kiss

It's too fast, it burnt out one of my ASUS ROG MARS cards. Wink


I have 3 spare 750ti's I am willing to sacrifice for the good fight.

Heck I will even up it to 5. Tongue
Jump to: